Re: CD.

[Brederlow <goswin.brederlow@student.uni-tuebingen.de>]

Michael Schmitz <SCHMITZ@LCBVAX.CCHEM.BERKELEY.EDU> writes:

> Hi,
> 
> >1. To see if it compiles. No point in selling a CD with Source that doesn't work.
> 
> Nice, but no one thinks you're responsible for the source in the first place.

We are sort of responsible for everything we put on the CD.

> >2. Optimisation (speed)
> 
> I hope you're not using optimization different from what the source maintainer
> used? There's a reason for not over-optimizing code in many places. But again, 

Nope, we just recompile the packages as they are. It's still faster
afterwards. The optimisation from egcs is better than gcc.

> not changing the optimization would mean you don't need to recompile (the fact
> that a binary package exists proves, at least to me, that the code compiles).

Sources should compile, but they don't more often that I like. Most
Packages that broke on the first pass missed dependencies and couldn't 
find include files. Thats my fault (or fault of missing sourcecode
dependencies). But some packages don't compile, which could be
Packages that don't exist for m68k in its newest version and are
therefore untested.

Another reason for recompiling is that many Packages give a lot of
warnings, some of which could be critical (like 'end of non-void
function reached'). Pointing out those warnings to the maintainer can
help to kill bugs.

> >> ecgs is, for all I know, not a stable compiler, and it's definitely not advised
> >> for e.g. kernel compilation. There's a reason why Debian sticks to gcc, I'm 
> >> sure.
> >
> >There is an egcs 1.0 release wich works fine. At the moment we have
> >egcs compiled kernels running on all the comps here but one. No
> >crashes so far. Compiling a kernel is a pretty good test for a
> >compiler and egcs did that just fine. (30 mins for a kernel with -O9)
> 
> I don't mind the speed of kernel compilation, I care about the stability of 
> the kernel. Linus warns against using egcs for kernel builds. 

Through egcs I was able to track some bugs in my code and one bug in
the kernel. That alone makes a reason to recompile with egcs.

May the Source be with you.
			Mrvn


--
TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to
debian-68k-request@lists.debian.org . 
Trouble?  e-mail to templin@bucknell.edu .